gpt-neo-125m: benchmark results
EleutherAI's GPT-Neo 125M, an open replication of the GPT-3 architecture trained on the Pile for 300 billion tokens, the smallest model in EleutherAI's original GPT-Neo release. Provider: EleutherAI. Released 2021-04-06. Access: Open.
Aggregate Rating 1120 ± 17, rank #10514 of 10916 rated models, from 80 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - High School Statistics | 45.83 | Accuracy (%) | 76.9 |
| MMLU-by-task - High School Physics | 31.13 | Accuracy (%) | 56.1 |
| MMLU-by-task - TruthfulQA MC2 | 45.58 | Accuracy (%) | 54 |
| MMLU-by-task - Professional Medicine | 44.85 | Accuracy (%) | 45.7 |
| MMLU-by-task - College Computer Science | 34 | Accuracy (%) | 39.6 |
| MMLU-by-task - Security Studies | 40 | Accuracy (%) | 39.1 |
| MMLU-by-task - TruthfulQA MC1 | 25.83 | Accuracy (%) | 38.2 |
| MMLU-by-task - College Physics | 22.55 | Accuracy (%) | 34.3 |
| MMLU-by-task - High School Geography | 35.86 | Accuracy (%) | 33.8 |
| MMLU-by-task - High School US History | 29.9 | Accuracy (%) | 33.8 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 45.58 | MC2 (%) (0-shot) | 31.4 |
| MMLU-by-task - High School Government And Politics | 35.75 | Accuracy (%) | 31.1 |
Interactive version: theaggregate.ai/model?slug=gpt-neo-125m · How It Works · Data refreshed daily, snapshot 2026-11-08.